Integration of a gps aided strapdown inertial navigation system for land vehicles

Abstract

The estimation accuracy of a low-cost inertial navigation system (INS) is limited by the accuracy of the used sensors and the imperfect mathematical modeling of the error sources. By fusing the INS data with GPS data, the errors can be bounded and the accuracy increases considerably. In this project, a low-cost in-house constructed inertial measurement unit (IMU) and an off-the-shelf GPS receiver are used for the data acquisition. The measurements are integrated with a loosely coupled GPS aided INS approach. For the assessment of the results, one data set with real data obtained from a field test is available. The tuning of the covariance matrices is a delicate adjustment and does not always provide convergence. Values for acceptable results could be found and two implementations of inertial navigation systems are compared. The use of nonholonomic constraints showed a dramatic increase in the accuracy. An analysis of the importance and influence of different IMU sensor errors provides a foundation for the modeling and inclusion of further error states in the extended Kalman filter.